LEVDA: Latent Ensemble Variational Data Assimilation via Differentiable Dynamics
A differentiable latent ensemble-variational smoother for joint state–parameter estimation from sparse, irregular observations.
I’m a Ph.D. candidate in Computer Science at Georgia Tech, advised by Peng Chen and Chao Zhang. I work on probabilistic machine learning and data assimilation for physical systems.
I’m also a graduate researcher at Oak Ridge National Laboratory, working on estimation and control for fluid flows. Previously, I worked on flood forecasting at Argonne National Laboratory.
Before Georgia Tech, I studied machine learning at Carnegie Mellon and Computer Science and Mathematics at Cornell, where I worked with Volodymyr Kuleshov.

I’m interested in estimating the state of a physical system when only a small part of it can be observed. I develop methods that combine generative models with learned dynamics to make forecasting and data assimilation faster.
Recent projects include LEVDA for joint state and parameter estimation, LD-EnSF for filtering with sparse observations, and CLDNet for high-resolution flood prediction.
A differentiable latent ensemble-variational smoother for joint state–parameter estimation from sparse, irregular observations.
Rainfall-driven latent neural dynamics and terrain-conditioned decoding for efficient, high-resolution basin-wide flood forecasts.
Latent dynamics and history-aware observation encoding accelerate forecasting and assimilation for complex dynamical systems.
Coupled variational autoencoders and ensemble score filtering for high-dimensional systems with highly sparse observations.
I enjoy cooking and gaming. I’ve put some photos of things I’ve made on the cooking page.